REPORT FROM THE
TASK FORCE ON
IMPROVING THE NON RESIDENT COLLABORATION ENVIRONMENT
7/15/94
In mid-February of this year the SLD spokesmen assembled a task force to
examine the difficulties faced by SLD collaborators who are not resident
at SLAC. The participants in the task force are:
Richard Dubois (SLAC)
Ray Frey (Oregon)
Tony Johnson (SLAC)
Mohan Kalelkar (Rutgers)
Steve Manly (Yale)
Ida Peruzzi (Frascati)
Shiro Suzuki (Nagoya)
Bob Wilson (Colorado State) -- Chair
The committee is formally called "The Task Force on Improving the Non
Resident Collaboration Environment". A primary focus of the task force was
to examine ways non resident collaborators may contribute more effectively
to SLD data analysis. Clearly, the lot of SLAC resident collaborators is
far from perfect, consequently it remains a natural priority of the
collaboration to provide a productive environment for this crucial
core effort. However, the role of this task force is to identify
obstacles of particular concern to off site people and to suggest actions
which address their particular problems.
Our first task was to determine just what the problems were. Members of
the task force polled colleagues at their institutions and held many
discussions with a wide variety of SLD collaborators. From this exercise
we generated a list which summarized the major points for discussion.
One thing became clear quickly, lack of information was a contributing
factor to much of the dissatisfaction. Quite often ignorance of existing
information and resources was at the root of the disgruntlement.
Another major effect of poor communication is the difficulty to maintain
the sense of community which is so important for effective collaboration.
Although communication among non-resident collaborators will never be as
easy as for those at SLAC, it is evident that the situation can be
improved with relatively little additional effort.
In this report we will present a list of suggestions for changes which,
in sum, we believe will substantially improve the ability of non
resident collaborators to contribute effectively to SLD operation and
analysis. Many of them we expect to be of short term utility,
essentially they are there to break the logjam which has built up of the
past few years. Some may be viewed as "obvious", barely worth writing
down. However, the task force felt that in some cases it was important
to emphasize some motherhood and apple-pie principles. The list represents a
distillation of many discussions and suggestions the task force has received
during the past few months. It is not intended to be all encompassing.
Undoubtedly, there are things we have missed and new issues will arise. The
first recommendation of the task force is that a standing committee be
established to monitor the implementation of these recommendations and to
act as a point of contact for issues as they arise.
We have divided the report into three main sections. Section I contains the task
force recommendations to the SLD spokesmen and the SLAC-based
collaborators, particularly those who form the core of the software
and analysis groups. Many suggestions for improvements were implemented
during the course of the task force investigations. To document some of this
activity we have included a section to chronicle important recommendations
which have already been acted upon. In Section III we have catalogued a set
of actions that a non resident group should take if they wish to make an
effective contribution to SLD analysis. Data from a questionnaire sent to all
institutional representatives provided valuable feedback to these
recommendations. A copy of the questionnaire and a summary of the replies is
given in Appendix I.
Section I: Recommendations to SLD Spokesmen and SLAC-Based
Collaborators
Computing and Data Access
- All collaborators should be strongly encouraged to migrate away from using
the IBM/VM system as soon as possible.
Discussion
Many collaborators still actively use the VM system, however, SLAC has
essentially stopped support for that system and the SLD software group no
longer has the manpower to support SLD software on that platform. For the
near term (1-2 years), the DEC/VMS operating system will the only system
fully supported for SLD analysis. The conversion of SLD software to unix
is not likely to provide a broadly supported analysis platform for SLD
for 1-2 years. Nonetheless, several groups are now using unix machines
effectively by generating specialized datasets in n-tuple format which
are then analyzed with tools such as PAW.
A netnews group:
slac.discuss.vmphaseout
has been established to discuss issues related to the VM phaseout.
A proposal has been prepared by a SLAC committee which addresses many of
the issues related to VM phase-out. It includes a schedule which effectively
terminates the use by SLD (and others) collaborators by September, 1995.
The document is
available on WWW.
- A non resident "SLD Computing Liaison" should be established to act as a
central point of contact for non resident collaborators. This person
should be familiar with SLD documentation and be able to answer "typical"
questions. The liaison would participate, by phone conference if
necessary, in the weekly software group meetings.
Discussion
Currently, non resident collaborators face long, frustrating delays when
trying to get answers to questions about SLD software. One reason
for this is the difficulty to locate (or even knowing the existence of)
the appropriate documentation. Another reason is that a few experts at
SLAC (e.g. Richard and Tony) receive so many e-mail requests for help
that it is inevitable that individual replies will not be addressed
quickly. On-site people can reinforce a request by a personal appearance
in the appropriate doorway, not an easy option for the non resident.
- Support for the transfer of SLD datasets to collaborating institutions
should be increased. A part-time data aide should be hired to perform
routine copying of datasets for distribution to offsite institutions.
Discussion
Several groups now have substantial VMS computing power. A key factor to
utilize it efficiently is access to SLD datasets. It should be
investigated whether off site institutions could help to support this
position.
- The collaboration spokesmen should solicit statements of computing needs
of collaborating groups. These should be assembled into a coherent
statement of the computing needs for the experiment as a whole.
Discussion
This document should be communicated to the appropriate funding agencies
which may use it as a basis to consider the funding requests of individual
institutions. This follows the model used for the construction of several HEP
experiments.
- The SLD software group should be allocated a separate budget to address
issues of offsite support.
Discussion
This will avoid some conflicts faced by the group in the current scheme.
A recent example of such a conflict is the choice between the purchase of more
hard disk for local systems or a tape stacker for use mostly by non resident
collaborators.
- The SLD software group should maintain a list of pending projects which
currently do not have an assigned executor, e.g. DUCSNEWS for the Alphas.
Discussion
This would enable collaborators to see what are the high priority needs
for SLD and to offer their services. Many projects would lend themselves
to offsite attention quite readily.
- The SLD physics working groups should maintain a list of tasks which
currently do not have an assigned executor, e.g. run a monte carlo to
determine the backgrounds to a process under study.
Discussion
It is often difficult for people wishing to get involved in an analysis
to determine where they can quickly make a contribution. A collaborator
at SLAC might begin attending working group meetings, speak with people
in the group and gradually become aware of the opportunities. It is more
difficult for someone who cannot easily visit SLAC to gather sufficient
momentum by this mechanism. It would be helpful if one could browse
through the "to do" lists that most analysis groups generate. This way
the remote collaborator could recognize a topic of interest and
the right scope before contacting the busy working group leaders.
- SLD should recommend to SLAC that SLUO be assigned a budget to
support X terminal access for visitors.
Discussion
A windowing environment (typically using the X protocol) has become a
standard for modern computer user access and much of SLD software is being
developed in that direction (SLD WWW pages and MidasWWW). SCS plans to phase
out support for ascii terminals such as the Ambassadors which are in common
use by visitors. Providing replacements is an important service that the
laboratory should provide for its users. SLD should strongly support the efforts
of SLUO to improve this service.
- Several institutions make use of the Micom switch for access to SLAC.
A request should be made to SCS to publish guidelines for alternative
modes of access.
Discussion
Access to SLAC computers for users with leased lines typically has been
through the MICOM switch. SCS has announced plans to discontinue this
service. These users need advice on the best ways to make use of these
expensive, but important (see item III.6) leased lines.
B. Meetings
- Televideo conferencing using the SLAC system should not be emphasized
for the near term.
Discussion
Currently, few non resident collaborators have convenient access to a
full-blown teleconference system. It is important that meetings which do
use this technology are also accessible with a more basic phone conference
link.
- The monthly SLD collaboration meeting should be held with at least one
way teleconferencing i.e. a remote collaborator should be able to
listen-in on the meeting.
Discussion
A minimal system would consist of two good quality microphones attached
to a phone, one located next to the chair of the meeting (usually the
spokesmen) the other near the projector (not too near the fan!).
- A speakerphone suitable for multi-way teleconference calls should be
purchased.
Discussion
A system which could accommodate up to 15 people at SLAC would service
most of the SLD meetings of interest to non resident collaborators (most
notably the physics analysis groups). A review of meeting times indicates
that one system will cover our needs, the Group C speakerphone would still
be available as a backup.
- A fax machine should be purchased for use in teleconference meetings.
Off-site users should also have one with convenient access.
Discussion
An obvious drawback of audio-only conferences is not being able to see
the transparencies. Faxing copies ahead of time is a partial solution
but quite often tranparencies are made at the last minute, or
are modified during the meeting. A fax machine immediately to hand
would improve the chances for off site people to have uptodate copies.
[Note: we have verified that one can fax directly from transparencies, but
don't use green or yellow!]
- The Orange room and at least one other room able to accommodate 10 to 15
people should be outfitted with TWO telephone lines. The Group A and C
meeting rooms and CL G104 (TAS's office) are good candidates.
Discussion
The conference phone and fax may then be used simultaneously in these
rooms. We presume that the Orange room will continue to be easily
available for SLD meetings, in particular the monthly collaboration
meeting.
- When requested by a collaborator, an effort should be made to hold
meetings in the SLAC videoconference room. These meetings should be advertised
at least one week in advance.
Discussion
Few collaborators have exclusive use of these systems and some advanced
booking is usually required (as it is at SLAC).
- Alternative televideo systems by the end of the year. A committee should
be formed to make a recommendation to collaborators for systems which
would make best use of this technology (similar to the recommendations
for computer systems given in the past). A SLAC resident SLD person
should be designated to participate in these investigations.
Discussion
The MBONE system is now in fairly common use for meetings ranging from
person-to-person, to full conference broadcast (such as the 1994
Computing in High Energy Physics conference). Several groups (including SLAC)
are investigating the MBONE system, however there is not enough experience to
recommend this route to SLD yet.
- Determination of the dates for SLD weeks and full collaboration meetings
should be done in consultation with the institutional representatives. An
effort should be made to position them close to other important meetings
at SLAC e.g. EPAC, DOE Review etc.
Discussion
Although it is clearly impossible to accommodate the busy schedules of
all collaborators, it would be useful to solicit comments on a preliminary
schedule before it is finalized.
- All major meetings of the Physics Working Groups should be scheduled to
occur during SLD weeks. A comprehensive calendar of meetings will be
prepared to facilitate this
(Now available!).
Published meeting times should be changed
only due to extraordinary circumstances. Schedule changes, particularly
those during SLD week, should be announced as soon as possible (via
DUCSNEWS, WWW, or e-mail).
Discussion
Non resident collaborators typically must plan travel dates at least one
to two weeks in advance to take advantage of discount air fares. Quite
often, due to teaching constraints, the trip will be for just a few days
so a change of meeting schedule can effectively invalidate that trip.
- Minutes of most SLD meetings should be posted electronically as soon as
possible after the meeting, preferably within one week. The minutes of
the physics working groups meetings would be particularly useful.
Discussion
This would also be of benefit to resident collaborators. A good example
is the daily SLC status report which is a very effective way for all
collaborators to keep in touch with the progress of the experiment.
- A copy of transparencies of meetings should be given to the SLD secretary
to keep a central file. In many cases this can be accomplished by simply
adding the secretary to existing distribution lists.
Discussion
If we succeed to have minutes of meetings readily available (i.e.
electronically) a collaborator can browse them easily and decide that the
transparencies would be useful. A request could then be made for copies of them
in the central file to be mailed or faxed. Storing all transparencies for
electronic access would be more convenient for the requestor but we feel
that the process would be too consuming of time and resources to be practical
for SLD.
General Communication
- An online list of collaborators should be kept which identifies their
area of expertise (loosely defined) and willingness to answer questions
with fairly rapid turnaround.
Discussion
A particular difficulty for non resident collaborators is lack of
opportunity to just wander down the corridor to find the answer to
nagging problems. The people on the list do not need to be part of the
official software group nor do they need to be at SLAC. Indeed, part of
the goal would be to lighten the burden on the traditional experts.
- The current DUCSNEWs system should be enhanced by a NETNEWS-DUCSNEWS
gateway, or possibly replaced by NETNEWS.
Discussion
NETNEWS feeds can be set up on local systems allowing more reliable and
faster access. NETNEWS is widely supported on many systems. The SLD
confer is no longer supported even on VM.
- The use of SLD Conference/NETNEWS as a forum for discussions
should be encouraged.
Discussion
Currently, the confer utility is used mostly for announcements. People
who frequent some of the USENET groups have seen how effective they can
be as a place for getting and giving advice [there is a lot of garbage
too, the discipline of the users makes the difference].
- The SLD electronic conferences list should be rationalized. Any
conference which has not received a new entry for more than 6 months
should be archived (in an easily accessible manner) and removed from the
public list (such as that presented by the WWW page).
Discussion
This process could occur during the migration to NETNEWS mentioned above.
- A list of SLD Notes and other internal documents should be made easily
accessible by computer. The location of the documents, preferably in
their electronic form, should be documented. Important documents, such as
physics papers, should be available electronically in their entirety i.e
including figures. An attempt should be made to provide at least the
text in electronic form for SLD Notes etc.
Discussion
For people at SLAC these documents are usually just a few corridors
away. Electronic access to them could save a great deal of time for those
who cannot visit SLAC frequently.
Section II: Recommendations Completed or In-Progress
A. Computing and Data Access
- Support for the transfer of SLD datasets to collaborating institutions
should be increased. The data copying (to 8mm tapes) process should be
simplified and made more robust.
Discussion
Two 8mm (2Gb/tape) drives have been moved to the Central Lab for easier
access. The procedure for tape copying has been simplified. Instructions
are
posted on WWW.
- SLD should provide some X terminal and workstation access for non
resident collaborators.
Discussion
The four VAX4000 will be re-located to the area currently occupied by the
color printer.
- Karen Heidenreich should be given discretion to install a "small"
number of archived tapes back into the silo upon request by a collaborator
Discussion
Chain-of-command request delays will be eliminated.
Meetings
- The schedule of all SLD meetings should be posted to DUCSNEWS and WWW.
Discussion
In progress. Kazuko is being instructed on the procedure to post items to
DUCSNEWS,
WWW and/or a new SLD Calendar conference.
- Cheaper alternatives for teleconferencing using workstations and the
internet for audio and video conferences are being investigated.
Discussion
The MBONE system is now in fairly common use for meetings ranging from
person-to-person, to full conference broadcast (such as the recent
Computing in High Energy Physics). Several groups (including SLAC) are
investigating the MBONE system.
General Communication
- Documentation, including programming examples, should be updated.
Outmoded examples and documents should be removed from circulation (e.g.
files removed from public SLD disks).
Discussion
This cleanup has been underway for a couple of months. Several help
additions have been made to the SLD WWW pages.
- A formal problem tracking system,
GNATS, has recently been instituted.
Discussion
This system should ensure that once a collaborator has reported a problem
it will not fall through the cracks.
- SLD "Paper Reading" should be scheduled at least one week in advance
and should occur in rooms equipped for teleconference.
Discussion
These readings are already quite well advertised and they follow the
guidelines laid out in the "Publications" document approved by the
SLD Council.
Section III: Recommendations for Non Resident Collaborators
In this section we present a list of recommendations, in no particular
order, which the task force believes are important to act on if
non resident collaborators wish to be effective members of the SLD
analysis effort.
- Attendance at SLD monthly meetings is an important means to keep in touch
with the activities of the collaboration. A visible presence also demonstrates
a commitment to the experiment and makes it easier for the resident experts to
justify the amount of time and effort they devote to supporting non resident
collaborators. If a trip to SLAC is not possible for each monthly meeting,
attending via teleconference should be a high priority.
- Graduate students/post docs. should be sent to SLAC for a substantial
period to get started on analysis i.e. more than just for shifts and
occasional SLD Weeks.
- Groups with VMS systems should install DUCS and assign a knowledgeable
person to monitor its use. [Contact Pieter Jacques at Rutgers:
JACQUES@RUTHEP, (908) 445-4431]. Those without VMS systems should
actively pursue either acquiring one or getting the best network access
possible e.g. hounding local and/or network administrators to improve
the connection to SLAC, or purchase of leased line etc. The task force
is recommending that the collaboration spokesmen play an active role
to support requests to funding agencies for these improvements.
- Analysis talks at collaboration meetings should be a high priority
goal.
- Network connections often cannot provide the low latency necessary to
make remote interactive logins an attractive way to work. Collaborators
are encouraged to use the networks in a way which does not require sub-
second response times, such as WWW and DUCS.
- Off-site groups with leased lines are advised to maintain them
(increased to 19.2 kbaud, if possible) until network access to SLAC
improves. Although leased lines are expensive to maintain, a poll of SLD
institutions indicates a general unhappiness with network access to SLAC.
The cause of these problems is multi-faceted and a moving target.
Typically, there are bottlenecks somewhere on one or more of the many
"hops" between institutions and SLAC. These can only be addressed by
institutions to their local networks and by informing interstate
networks, such as ESNET, of problems. On occasion the problem appears to
be within the SLAC system, this is also a concern of on-site users. If this
seems to be the case, please let SCS know, they are willing to help if they
know there is a problem.
- Monitor the
SLD Home Page
on the World Wide Web frequently since it
has become the primary source of information within SLD. SLAC also
intends to use WWW heavily to address some of the issues of the move
from centralized computing (VM) to distributed systems (unix, pcs, etc.).
- It is important for the people at SLAC, who are asked to do much of
the service work, to see that their efforts are contributing to a
stronger physics program for SLD. With this in mind, it is particularly
important for non resident collaborators to make a substantial effort to
give frequent updates of their work. For example, they should regularly
attempt to present their work at the physics working group meetings
and to make the effort to schedule teleconference meetings.
- Offer to serve as experts to help other remote collaborators e.g.
setting up DUCS (as Rutgers has done), setting up the SGI, etc.
- Check the "task lists" of the analysis working groups. Although this
is no substitute for actively participating in a group it may be a good
way to identify topics which may interest you and which you feel you
can pursue somewhat independently of the SLAC based group.
- Volunteer to address some of the items in the list of outstanding
offline tasks (the "Offline Job Jar") e.g. DUCSNEWS on the Alpha,
"clean-up" of the documentation, etc.
- Volunteering for the "Offline Shifts" is another good way to keep
in close contact with the experiment, perform an important task which
maintains data quality, and demonstrate a willingness to contribute
to the general infrastructure of SLD analysis.
- Remember to report problems you have. It is impossible for the SLAC
people to anticipate everything that could go wrong or to make all
documentation perfectly clear. They need feedback. For things that do not
need immediate attention (IDA bugs for example) the
GNATS problem
tracking system should be used.
For faster attention, for smaller problems, consulting the list of
volunteer experts (which we hope to establish soon) is the better
approach. Posting questions, and their subsequent solution, to the
SLD conferences (or their replacements) will help to educate all of us too.
APPENDIX I
TO: SLD Council Representatives
SUBJECT: Questionnaire related to SLD offsite analysis efforts
FROM: SLD Task Force on Improving the Non-Resident Collaboration Environment.
REQUESTED REPLY DATE: Thursday, June 9th 1994
In an effort to improve the effectiveness of SLD collaborators wishing to be
involved in analysis we have prepared the following questionnaire. Your
answers provide very important input which will help us to make best use of
our limited resources (remember that this includes both hardware AND
peopleware). As the representative to the SLD council we are relying on you
to solicit as much information as possible from your colleagues and to return
this form to us in time for it to be useful in the upcoming run. You are
encouraged to add any other comments you think will help us at the bottom of
the form.
Please complete this questionnaire and return it to BOBW@SLAC.STANFORD.EDU by
Thursday, June 9th 1994. Thanks.
1. PERSONNEL
How many of each of the following people (not resident at SLAC) do you
have ACTIVELY involved in SLD analysis:
grad. students 9
post docs 3.5
faculty 7_
How many of each of the following people (not resident at SLAC) do
estimate you would have actively involved in SLD analysis if it were
easier to do so:
grad. students 15.8
post docs 9.5
faculty 11
2. COMPUTING RESOURCES
Home institution computing resources available for SLD activity (list
even if not currently in use):
Operating system(s): a. VMS______ , b. UNIX______ , c. Some IBM _
Approx. MFLOPS: a. 250 - 300 , b. 100 - 200_ , c. and SGI __
Fraction for SLD use: a. _________ , b. __________ , c. __________
Disk space (Gbytes): a. 35 _____ , b. 12 ______ , c. __________
Tape drives (type): a. 13 8-mm__ , b. 2 4-mm ___ , c. __________
System Manager (Y/N): a. < 2/3 yes , b. __________ , c. __________
Do you have software expertise which could be a resource to the
collaboration (e.g. a system manager willing to help other groups
install new software)? ___
At what level? _____ Little excess capacity, most none
3. REMOTE MEETINGS
Do you have a televideo system? 3Y 6N
System: _________________________________________
Accessibility: ___________________(good, limited, none)
Do you think televideo would be an effective tool for SLD? most No
Do you know what MBONE is? 4Y 4N
(You may want to ignore the next few questions if you answered NO)
Do you have conveniently available MBONE capable workstations (SUN, SGI,
HP)? _ 2Y ________________________________ (# available and type)
Are they equipped with a video camera? NO
Have you used MBONE? DEMO
Do you think it would be an effective tool for SLD? UNSURE
Do you have audio-only phone conferencing facility? 5N 4 Y
Do you think it would be an effective tool for SLD? 4Y 2N
Have you tried a conference call using just a speaker-phone? 7Y 2N
Do you think it would be an effective tool for SLD? 4Y 2N
Which meetings would you be most likely to attend remotely?
Monthly SLD, Physics Working Groups, CRID, 8:30 mtgs ___
4. OTHER COMMUNICATION
DUCS
Do you use DUCS regularly? 2Y 7N
Do you find DUCS reliable? 2N 1Y 3?
Do you think that DUCS is sufficient to allow offsite analysis?2N 1Y
1?
World Wide Web (WWW)
Do you use WWW regularly? 6Y 2N
Do you know that SLD has an extensive WWW page? 9Y
{Look at http://slacvx.slac.stanford.edu/SLDWWW/000000/sld.html}
Do you find the WWW useful? 9Y . Do you find SLDŐs WWW page useful? 7Y 1N
Do you think that WWW is a good way for SLD to communicate? 7Y 2N
{Please try the SLD Home Page first!}
5. SLAC COMPUTING ACCESS
Which SLAC computers do you use most (SLACVM, SLACVX, ALPHAs, UNIX)?
6 SLACVX, 6 SLACVM, 1 ALPHA _____________
Are you satisfied with the computing available to you? 3N 4Y 2?
Are you satisfied with your network access to SLAC computers?3N 4Y
Do you maintain a leased line to SLAC? 3Y 4N What speed? 9600
Do you use SLAC computers mostly for analysis or communications purposes?
(e.g. for e-mail, SLDnews etc. or running analysis jobs, IDA etc.)
____________________ ALL OF THE ABOVE ___________________
6. DATA ACCESS
For effective analysis at your home institution how important would it be
to have the following datasets transmitted to YOUR SITE (i.e. the dataset
available to you on local disks or tapes)?
{See the SLD Primer on WWW
http://slacvx.slac.stanford.edu/SLDWWW/000000/sldprimer.html
for a complete description of the datasets and current ways to access them}
[ Scale: 1-crucial, 2-important, 3-nice, 4-not needed, 5-waste of time
and resources ]
ACQ/RAW 4.0
PASS1 3.5
PASS2 3.1
MonteCarlo 2.2
miniDST 2.3
microDST 2.0
Do you have 8mm tape drives available at your home location for your use
for data transfer? 8Y 1N
Do you have 4mm or other tape drives available for your use for data
transfer? 2Y 5N
Would you be willing to purchase an appropriate system if other issues
related to offsite analysis were addressed? 5Y 2N ALL MENTIONED
$$ PROBLEMS
7. GENERAL
What two things (that might be feasibly changed) can we change that you
think would increase your productivity and effectiveness on SLD?
1._____________________________________________________________________
2._____________________________________________________________________
8. COMMENTS
Please add any other comments you would care to make: